1. Field of the Invention
The present invention relates generally to semiconductor light sources such as LEDs and VCSELs, and more particularly to a strained layer semiconductor laser having an emission wavelength of at least 1.3 .mu.m.
2. Description of the Prior Art
Vertical-Cavity Surface-Emitting Lasers (VCSELs), Edge Emitting Lasers (EELs) or Light Emitting Diodes (LEDs) are becoming increasingly important for a wide variety of applications including optical interconnection of integrated circuits, optical computing systems, optical recording and readout systems, and telecommunications. Vertically emitting devices have many advantages over edge-emitting devices, including the possibility for wafer scale fabrication and testing, and the possibility of forming two-dimensional arrays of the vertically emitting devices. The circular nature of the light output beams from these devices also make them ideally suited for coupling into optical fibers as in optical interconnects or other optical systems for integrated circuits and other applications.
For high-speed optical fiber communications, laser or LED emission wavelengths in the 1.3 .mu.m through 1.55 .mu.m region are desired. Standard silica fiber has zero dispersion near 1.3 .mu.m and has a minimum loss near 1.55 .mu.m. The need for semiconductor lasers emitting in this wavelength region has spawned worldwide development of such lasers. Group III-V semiconductors which emit light in the 1.3 through 1.55 .mu.m region have lattice constants which are more closely matched to InP than to other binary III-V semiconductor substrates, for example, GaAs. Thus, essentially all commercial emitting lasers emitting at 1.3 through 1.55 .mu.m are grown on InP substrates. These lasers are edge-emitting lasers which, unlike VCSELs, do not require high-reflectivity Distributed Bragg Reflectors (DBRs) to form their optical cavities.
Unfortunately, it has proven difficult to produce effective DBRs on InP substrates. The available materials which lattice match InP have produced mirrors which are extremely thick and lossy and have thus not resulted in efficient VCSELs.
VCSELs or Surface Emitting Lasers SELs whose current flow is controlled by lateral oxidation processes have show the best performances of any VCSELs in terms of low threshold current, high efficiency, and high speed. All such "oxide VCSELs" have been fabricated using AlAs or AlGaAs layers which were grown on GaAs substrates and later oxidized. Thus, one would want to utilize a VCSEL structure such as is disclosed in U.S. Pat. No. 5,493,577, by Choquette et al. This VCSEL has the advantages of: (1) reduced mode hopping; (2) being temperature stable, and (3) testable in a modified silicon wafer tester. Unfortunately, this VCSEL structure will have an emission wavelength between 600 and 1,000 mn and thus falls short of the desired 1.3 .mu.m emission wavelength. Due to the availability of well-behaved oxidizable materials which may be grown on GaAs substrates and the straightforward capability of producing efficient high-reflectivity DBRs on GaAs substrates, when manufacturing VCSELs it is highly desirable to grow them on GaAs substrates.
The salient components of a VCSEL typically include two DBRs, and between them, a spacer which contains an active region having a length emitting material. The DBRs and active region form an optical cavity characterized by a cavity resonance at a resonant wavelength corresponding to a resonant photon energy. It has become a practice in the operation of VCSELs to detune the optical cavity to energies at about 25 meV lower than the peak transition energy by appropriate DBR spacing. Such "gain offset" is used to advantage in reducing temperature sensitivity. This produces an emission wavelength which is appreciably longer than the peak transition wavelength. This practice, while inadequate in itself for increasing emission wavelength to 1.3 .mu.m from material grown pseudomorphically on GaAs substrates, does measurably increase emission wavelength. Even if this technique was incorporated with the teachings of the prior art, one would fall short of the desired 1.3 .mu.m emission wavelength.
While epitaxial growth of slightly-lattice-mismatched materials is undertaken routinely, materials which emit in the 1.3 .mu.m through 1.55 .mu.m region have lattice constants sufficiently removed from that of GaAs to make pseudomorphic epitaxial growth problematic. In this context, "pseudomorphic" means having a sufficiently low density of misfit dislocations such that lasers may be produced which have reasonably long lifetimes. The problems have been sufficiently great to cause researchers to abandon such efforts and resort to less desirable hybrid approaches to producing 1.31 .mu.m through 1.55 .mu.m VCSELs.
Thus, the production of VCSELs emitting at 1.3 through 1.55 .mu.m wavelengths has been inhibited by either of two problems. The problems result from the fact that VCSELs require laser-quality active materials and high-reflectivity DBR mirrors. These two problems are:
(1) when InP substrates are used, growth of the light emitting active material is straightforward, but production of efficient DBRs is difficult and has not been effective; and PA1 (2) when GaAs substrates are used, DBR production is straightforward, but efforts to grow laser-quality active material have been unsuccessful.
The following is a summary of the prior approaches which are relevant to the problem of producing 1.3 though 1.55 .mu.m VCSELs.
A 1.3 .mu.m edge-emitting laser grown on a GaAs substrate was reported by Omura et al., in an article entitled "Low Threshold Current 1.3 .mu.m GaInAsP Lasers Grown on GaAs Substrates," Electronics Letters, vol. 25, pp. 1718-1719, Dec. 7, 1989. The structure comprises a layer having a high density of misfit dislocations, on top of which were grown thick layers of materials having lattice constants close to that of InP. Such lasers exhibit very poor reliability due to the misfit dislocations. Furthermore, this structure does not readily lend itself to integration with DBR mirrors.
The use of a layer having high-density misfit dislocations was also reported by Melman et al., in an article entitled "InGaAs/GaAs Strained Quantum Wells with a 1.3 .mu.m Band Edge at Room Temperature," Applied Physics Letters, vol. 55, pp. 1436-1438, Oct. 2, 1989. The article states that pseudomorphic, i.e., nearly misfit dislocation free, growth of 1.3 .mu.m emitting material is not possible with GaAs barriers, i.e., GaAs substrates. This conclusion prompted the approach to incorporate a layer having a high density of misfit dislocations.
A 1.1 .mu.m emitting laser is reported in Waters et al., in an article entitled "Viable Strained Layer Laser at .lambda.=1100 nm," Journal of Applied Physics, vol. 67, pp. 1132-1134, Jan. 15, 1990. The laser utilized a single quantum well comprising In.sub.0.45 Ga.sub.0.55 As strained semiconductor material which has a greater thickness than its predicted critical thickness. Reliability tests are presented for 4000 hours of testing. To our knowledge, these are the longest-wavelength lasers produced on GaAs substrates which have survived such testing. In this article, even Waters recognizes the difficulty of creating a reliable device having an active region over the respective CT for the semiconductor material in the active region.
A strained quantum well emitting at 1.31 .mu.m is reported by Roan and Chang in an article entitled "Long-Wavelength (1.3 .mu.m) Luminescence in InGaAs Strained Quantum-Well Structures grown on GaAs," Applied Physics Letters, vol. 59, pp. 2688-2690, Nov. 18, 1991. The quantum well was a short-period superlattice comprising alternating monolayers of InAs and GaAs. However, the quantum well had a thickness well above (1.78 times) the critical thickness, above which high densities of misfit dislocations exist. Thus, the structure is not viable for long-lived lasers and no lasers were produced from such a structure.
A compromise between GaAs and InP substrates is reported by Sahoji et al., in an article entitled "Fabrication of In.sub.0.25 Ga.sub.0.75 As/InGaAsP Strained SQW Lasers on In.sub.0.05 Ga.sub.0.95 As Ternary Substrate," IEEE Photonic Technology Letters, vol. 6, no. 10, pp. 1170-1172, Oct. 10, 1994. An In.sub.0.05 Ga.sub.0.95 As ternary substrate was utilized which has a lattice constant intermediate between those of GaAs and InP. The In concentration of the substrate was 5% of the group III material (2.5% of the total material) and the laser emitted at 1.3 .mu.m. The authors indicate that 1.3 .mu.m lasers will require an InGaAs substrate having about 25% or more In for the group-III material. Ternary substrates are unlikely to approach the availability, size and price of binary substrates such as GaAs.
James Coleman, in his book entitled "Quantum Well Lasers," edited by Peter Zory, London, Academic Press, pp. 372-413, 1993, discusses the concept of critical thickness in strained layers lasers which utilize In.sub.y Ga.sub.1-y As. As may be seen in FIG. 4 of this reference, as the composition of In increases, i.e., y approaches 0.5, the critical thickness drops dramatically. Turning now to FIG. 10 of this reference, it may be seen that Coleman has demonstrated that as the In concentration increases, the peak transition wavelength increases in a sub-linear fashion. As the In concentration approaches 0.5 the peak transition wavelength approaches about 1.20 .mu.m. If one was to extrapolate information from this graph for In concentrations greater than or equal to 0.5, one would come to the clear conclusion that a peak transition wavelength of 1.3 .mu.m is not obtainable while maintaining the In.sub.y Ga.sub.1-y As layer within the critical thickness. Thus, while Coleman does provide a valuable teaching, he is unable to reach a 1.31 .mu.m peak transition wavelength.
The issue of strain compensation to increase the number of strained quantum wells which may be grown without misfit dislocations is frequently used in the art and is described by Zhang and Ovtchinnikov in an article entitled "Strain-compensated InGaAs/GaAsP/GaInAsP/GaInP Quantum Well Lasers (.lambda..about.0.98 .mu.m) Grown by Gas-Source Molecular Beam Epitaxy," Applied Physics Letters, vol. 62, pp. 1644-1646, 1993. The reader is also referred to U.S. Pat. No. 5,381,434 by Bhat and Zah.
The advantages of incorporating strain into the active region of a semiconductor laser were described by Yablonovitch in U.S. Pat. No. 4,804,639. Yablonovitch discloses active regions of In.sub.y Ga.sub.1-y. As grown on GaAs substrates, typically with y.about.0.5, and having a thickness preferably less than 100 .ANG.. He suggests the possibility of "the addition of counter-strain layers of GaP on either side of the active strained layer," but does not pursue this possibility. He goes on to perform numerical evaluations based on "an assumed set of numerical coefficients which are thought to be representative of a quaternary semiconductor with a band edge near the 1.5 .mu.m wavelength." The material is further assumed to have a strain of 3.7% and a thickness of 100 .ANG. which was thought to be "probably the maximum permissible thickness for such a high strain." This strain for y=0.5 is calculated to be less than .about.40 .ANG.. Thus, although a &gt;1.3 .mu.m emitting laser utilizing strained InGaAs on GaAs is indirectly suggested, no actual structure is specified and the parameters are not realistic.
In U.S. Pat. No. 5,060,030, Hoke describes improvements in electron mobility and electron saturation for use in high-electron-mobility transistors (HEMTs). He describes the use of strain compensation to increase the thickness or In concentration "by approximately a factor of two" in a strained InGaAs layer grown on GaAs.
A strain-compensated heterostructure laser diode is described by Buchan et al. in U.S. Pat. No. 5,373,166. Buchan describes graded structures in the compressive and tensile strained quaternary layers with the strain magnitudes less than 1%. The thicknesses of the layers described are less than their conventional critical thicknesses.
Vawter et al., in an article entitled "Useful Design Relationships for the Engineering of Thermodynamically Stable Strained-layer Structures," Journal of Applied Physics, vol. 65, pp. 4769-4773, 1989, describes approaches for engineering dislocation-free strained-layer structures. The article includes a methodology for calculating the "critical thickness" of structures comprising layers of differing lattice constants. The methodology is based upon a "reduced effective strain" which is the sum of the strain-thickness products of all the layers divided by the total thickness of the layers. Based upon this "reduced effective strain," the "critical thickness" for the structures is then calculated from a critical thickness criterion, e.g., that introduced by Matthews and Blakeslee.
Asahi et al., in an article entitled "New III-V Compound Semiconductors T1InGaP . . . " Japanese Journal of Applied Physics, describes the inclusion of the group-III element thallium (T1) in III-V semiconductors for long-wavelength emission. Most of the discussion focuses on lasers emitting at wavelengths greater than 2 .mu.m on InP substrates, but it is stated that T1GaP lattice-matched to GaAs substrate has a bandgap emission of about 1.24 .mu.m. The extreme toxicity and hazardous nature of T1, even after epitaxial growth is performed, makes it undesirable as a manufacturing material.
Very recently, it has been shown that adding nitrogen to InGaAs, actually decreases the peak transition energy and thereby increases the peak transition wavelength as described by Kondow et al., in an article entitled "GaInNAs: A Novel Material for Long-Wavelength-Range Laser Diodes with Excellent High-Temperature Performance," Jpn. J. Appl. Phys., vol. 35, pp. 1273-1275, February 1996. The report suggests that it is difficult to grow high quality InGaAsN with very much N. A room temperature photo-luminescence spectrum of a 70 .ANG. thick In.sub.y Ga.sub.1-y As.sub.1-v N.sub.v /GaAs quantum well showed significant broadening even with only less than 1% N concentration for the group V semiconductor element. This corresponds to a value of v being less than 0.01. The peak transition wavelength of this semiconductor was 1.23 .mu.m. In the report, the authors state that their plan is to reach a 1.3 .mu.m device by increasing the N concentration to 1% while maintaining a 30% In concentration.
A hybrid approach to address the dual problem described earlier has been reported by Margalit et al., in an article entitled "Laterally Oxidized Long Wavelength CW Vertical-Cavity Lasers," Applied Physics Letters, vol. 69, pp. 471-472, Jul. 22, 1996. In this work, two DBRs are grown on two separate GaAs substrates, while the active material is grown on a third substrate which comprises InP. The active material comprises seven compressively strained InGaAsP quantum wells clad by 300 nm of InP on each side. To assemble these materials, two processes are performed, each including the thermal fusion of two wafers and removal of one substrate. Then the resulting structure is processed by standard VCSEL processing methods.
Since VCSELs are presently the subject of intense research and development, a great deal of results and advancements are published periodically. The following is a list of documents which are relevant to the problem of extending emission wavelengths of semiconductor lasers or of producing 1.3 .mu.m through 1.55 .mu.m VCSELs.
Fisher et al., "Pulsed Electrical Operation of 1.5 .mu.m Vertical-Cavity Surface Emitting Lasers," IEEE Photonics Technology Letters, Vol. 7, No. 6, pp. 608-609, Jun. 6, 1995. PA0 Uchiyama et al., "Low Threshold Room Temperature Continuous Wave Operation of 1.3 .mu.m GaInAsP/InP Strained Layer Multiquantum Well Surface Emitting Laser," Electronics Letters, vol. 32, no. 11, pp. 1011-1013, May 23, 1996. PA0 Hasenberg, "Linear Optical Properties of Quantum Wells Composed of All-Binary InAs/GaAs Short-Period Strained-Layer Superlattices," Applied Physics Letters., vol. 58, no. 9, pp. 937-939, Mar. 4, 1991. PA0 Fukunaga et al., "Reliable Operation of Strain-Compensated 1.06 .mu.m InGaAs/InGaAsP/GaAs Single Quantum Well Lasers," Applied Physics Letters., vol. 69, no. 2, pp. 248-250, Jul. 8, 1996. PA0 Kondow et al., "Gas-Source Molecular Beam Epitaxy of GaN.sub.x As.sub.1-x Using a N Radical as the N Source," Jpn. J. Appl. Phys., vol. 33, pp. 1056-1058, Aug. 1, 1994. PA0 Shimomura et al., "Improved Reflectivity of AlPSb/GaPSb Bragg Reflector for 1.55 .mu.m Wavelength," Electronics Letters, vol. 30, no. 25, pp. 2138-2139, Dec. 8, 1994. PA0 Blum et al., "Wet Thermal Oxidation of AlAsSb Lattice Matched to InP for Optoelectronic Applications," Applied Physics Letters., vol. 68, no. 22, pp. 3129-3131, May 27, 1996. PA0 Mirin, R. P., "1.3 .mu.m Photoluminescence From InGaAs quantum dots on GaAs," Applied Physics Letters., vol. 67, no. 25, pp. 3795-3797, Dec. 18, 1995.
Thus, although the prior art therefore describes a variety of techniques useful in forming long-wavelength lasers on GaAs substrates, it fails to provide any specific example of a viable such structure, nor does it provide any range of parameters within which viable such structures may be fabricated, nor does it teach the construction of a viable such structure. Some references suggest the possibility of 1.31 .mu.m lasers on GaAs substrates, but provide unrealistic parameters and are several years old or more.